Azure speech service error

Question

Azure speech service error

Sachin Raorane 20

We are using the Azure Speech Service SDK in our application for speech-to-text conversion.

We are currently using the Standard tier, but it occasionally fails with the following error: 'Connection was closed by the remote host. Error code: 1007. Error details: Quota exceeded.'

What could be causing this issue? We plan to use this service at a production level and need help resolving this.

Additionally, we would like to learn more about this service in depth.

Sachin Raorane 20 Reputation points

2026-06-18T09:54:18.4766667+00:00
hello @Anshika Varshney

Thank you for reply

as mentioned above

The Azure region of your Speech resource: East US

Whether you're using Speech-to-Text, Text-to-Speech, or another Speech feature : Speech-to-Text

The approximate number of concurrent requests when the issue occurs : 36 total calls (at 12.32 PM IST)

Whether the error is persistent or only happens during traffic spikes : need to check

I have one more query how many users can use this service in parallel as we want to use it in production & what is difference between azure Microsoft foundry and Azure AI foundry.
Anshika Varshney 14,270 Reputation points Microsoft External Staff Moderator

2026-06-23T19:23:52.0733333+00:00
Hello @Sachin Raorane

Thank you for sharing the details.

Regarding scalability, the number of users that can use Azure Speech-to-Text in parallel depends on your pricing tier and workload pattern rather than a fixed user count. For Standard (S0) Speech resources, the default limit for real-time Speech-to-Text is 100 concurrent requests per resource, and this limit can be increased through a quota request if required. Since you mentioned approximately 36 concurrent calls, that is generally within the default limit. [learn.microsoft.com],

For production deployments, it is recommended to:

Monitor for throttling (HTTP 429) and implement retry logic.

Load test your application using expected production traffic.

Request a quota increase in advance if you expect peak concurrency to exceed the default limits.

Regarding your question:

Microsoft Foundry is Microsoft's unified platform for building, deploying, and managing AI applications and agents. It brings together models, agents, tools, governance, monitoring, and enterprise controls in a single experience. [learn.microsoft.com], [Azure Spee...soft Azure | Azure.Microsoft.com]

Azure AI Foundry was the previous name of this platform. Microsoft has rebranded and consolidated the experience under Microsoft Foundry, while services such as Azure Speech are now part of Foundry Tools. The underlying AI capabilities remain available, but they are now organized within the broader Foundry ecosystem. [Foundry To...soft Azure | Azure.Microsoft.com]

Could you also confirm whether the Speech Service errors occur consistently or only during peak traffic periods? This will help determine whether the issue is related to throttling, service limits, or another cause.

I Hope this helps. Do let me know if you have any further queries.

Thankyou!
Anshika Varshney 14,270 Reputation points Microsoft External Staff Moderator

2026-06-24T15:06:41.3766667+00:00

Hello @Sachin Raorane

Please don't forget to close up the thread here by upvoting and accept my answer as an answer if it is helpful.

Answer accepted by question author

Sina Salam 30,566 Volunteer Moderator

Hello Sachin Raorane,

Welcome to the Microsoft Q&A and thank you for posting your questions here.

I understand that you're having Azure speech service error.

For Standard S0 real-time Speech-to-Text, the default concurrent request limit is 100 per Speech resource. Since you reported only 36 calls, 36 alone should not exceed the default limit.

Therefore, the actual issue is one of these:

Your application is leaving Speech recognizer sessions open,
More workloads are sharing the same Speech resource,
The app is using a different resource/tier/region than expected, or the quota/tier change has not propagated.

My advice to you:

Implement a strict client-side concurrency limit of 70–80 active recognizers,
Explicitly stop and dispose every recognizer, log SDK CancellationDetails including ErrorCode, ErrorDetails, and session ID, and verify metrics against the exact East US Speech resource.
If production concurrency must exceed 100, raise an Azure quota request for real-time Speech-to-Text concurrent requests on that Speech resource.

This will guide you for both mitigation and root cause.

I hope this is helpful! Do not hesitate to let me know if you have any other questions, steps or clarifications.

Please don't forget to close up the thread here by upvoting and accept it as an answer if it is helpful.

0 comments

1 additional answer

Your answer

Sachin Raorane 20 Reputation points

2026-06-18T09:54:18.4766667+00:00

hello @Anshika Varshney

Thank you for reply

as mentioned above

The Azure region of your Speech resource: East US

Whether you're using Speech-to-Text, Text-to-Speech, or another Speech feature : Speech-to-Text

The approximate number of concurrent requests when the issue occurs : 36 total calls (at 12.32 PM IST)

Whether the error is persistent or only happens during traffic spikes : need to check

I have one more query how many users can use this service in parallel as we want to use it in production & what is difference between azure Microsoft foundry and Azure AI foundry.
Anshika Varshney 14,270 Reputation points Microsoft External Staff Moderator

2026-06-23T19:23:52.0733333+00:00

Hello @Sachin Raorane

Thank you for sharing the details.

Regarding scalability, the number of users that can use Azure Speech-to-Text in parallel depends on your pricing tier and workload pattern rather than a fixed user count. For Standard (S0) Speech resources, the default limit for real-time Speech-to-Text is 100 concurrent requests per resource, and this limit can be increased through a quota request if required. Since you mentioned approximately 36 concurrent calls, that is generally within the default limit. [learn.microsoft.com],

For production deployments, it is recommended to:

Monitor for throttling (HTTP 429) and implement retry logic.

Load test your application using expected production traffic.

Request a quota increase in advance if you expect peak concurrency to exceed the default limits.

Regarding your question:

Microsoft Foundry is Microsoft's unified platform for building, deploying, and managing AI applications and agents. It brings together models, agents, tools, governance, monitoring, and enterprise controls in a single experience. [learn.microsoft.com], [Azure Spee...soft Azure | Azure.Microsoft.com]

Azure AI Foundry was the previous name of this platform. Microsoft has rebranded and consolidated the experience under Microsoft Foundry, while services such as Azure Speech are now part of Foundry Tools. The underlying AI capabilities remain available, but they are now organized within the broader Foundry ecosystem. [Foundry To...soft Azure | Azure.Microsoft.com]

Could you also confirm whether the Speech Service errors occur consistently or only during peak traffic periods? This will help determine whether the issue is related to throttling, service limits, or another cause.

I Hope this helps. Do let me know if you have any further queries.

Thankyou!
Anshika Varshney 14,270 Reputation points Microsoft External Staff Moderator

2026-06-24T15:06:41.3766667+00:00

Hello @Sachin Raorane

Please don't forget to close up the thread here by upvoting and accept my answer as an answer if it is helpful.

Answer 1

Hello @Sachin Raorane

Thank you for sharing the error details.

The key part of the error appears to be:

"Quota exceeded" (WebSocket close code 1007)

This typically indicates that the Speech service is rejecting requests because a service limit or quota has been reached. Even on the Standard tier, throttling can occur if request volume, concurrency, or transaction rates exceed the allocated capacity.

Things to check

Review Speech resource metrics
- Monitor request volume, concurrent requests, and error rates in Azure Monitor.
- Check whether the failures correlate with traffic spikes or peak usage periods.
Verify quota and limits
- Review the quotas applicable to your Speech resource and region.
- If your workload has recently increased, you may be hitting throughput limits intermittently.
Check Azure Service Health
- Review Azure Service Health and Azure Status for any incidents affecting Azure Speech Services in your region during the timeframe when the errors occurred.
Implement retry logic
- If you are not already doing so, consider implementing retries with exponential backoff for transient failures and throttling scenarios.

Could you please share the below details:

The Azure region of your Speech resource
Whether you're using Speech-to-Text, Text-to-Speech, or another Speech feature
The approximate number of concurrent requests when the issue occurs
Whether the error is persistent or only happens during traffic spikes

This information will help determine whether the issue is related to quota enforcement, workload patterns, or a service-side issue.

I hope this helps. Let us know the additional details, so that we can help investigate further.

Thankyou!