Azure Communication Services real time transcript failed to start

Question

Azure Communication Services real time transcript failed to start

Daniel Li 0

Hello Azure Support Team,

I am experiencing an issue with Azure Communication Services real-time transcription.

Environment:

Azure.Communication.CallAutomation Version = 1.5.0
Azure.Communication.Rooms Version = 1.2.0
Speech service configured in Azure
Role assignment: Cognitive Services User assigned from ACS to the Speech service and resource group

Code snippet:

           try
           {
               CallAutomationClient callAutomationClient = new CallAutomationClient(this.connectionString);

               Uri websocketUrl = new Uri(this.websocketUri);

               ConnectCallOptions connectCallOptions = new ConnectCallOptions(new RoomCallLocator(roomId), new Uri(this.callbackUri))
               {
                   MediaStreamingOptions = new MediaStreamingOptions(MediaStreamingAudioChannel.Unmixed)
                   {
                       TransportUri = websocketUrl,
                       StartMediaStreaming = true,
                   },
                   CallIntelligenceOptions = new CallIntelligenceOptions()
                   {
                       CognitiveServicesEndpoint = new Uri("https://****.api.cognitive.microsoft.com/")
                   },
                   TranscriptionOptions = new TranscriptionOptions(locale: "en-US")                    {
                       TransportUri = websocketUrl,
                       StartTranscription = true,
                   }
               };

               ConnectCallResult response = await callAutomationClient.ConnectCallAsync(connectCallOptions);

               return Ok(new
               {
                   serverCallId = response.CallConnectionProperties.ServerCallId,
                   callConnectionId = response.CallConnectionProperties.CallConnectionId,
               });
           }
           catch (Exception ex)
           {
               return BadRequest(new
               {
                   error = ex.Message
               });
           }

Issue:

I can successfully receive TranscriptionMetadata.
However, I do not receive any TranscriptionData.
The transcription status remains not active.

Request:

Could you please help investigate why the transcription does not start, even though metadata is received? Is there any additional configuration required for ACS real-time transcription with Speech service?

I appreciate your support.

ref:

https://learn.microsoft.com/en-us/azure/communication-services/how-tos/call-automation/real-time-transcription-tutorial?source=recommendations&pivots=programming-language-csharp

https://learn.microsoft.com/en-us/azure/communication-services/concepts/call-automation/azure-communication-services-azure-cognitive-services-integration

Vimal Lalani 2,560 Reputation points Microsoft External Staff Moderator

2025-12-18T20:43:27.4333333+00:00

Hi @Daniel Li

Just following up to see whether the above information helped address your question. Please feel free to post back if you need any further assistance.

1 answer

Your answer

Vimal Lalani 2,560 Reputation points Microsoft External Staff Moderator

2025-12-18T20:43:27.4333333+00:00

Hi @Daniel Li

Just following up to see whether the above information helped address your question. Please feel free to post back if you need any further assistance.

Answer 1

Vimal Lalani 2,560 Microsoft External Staff Moderator

Hi @Daniel Li

Thank you for posting your question on Microsoft Q&A.

Based on the details shared, the issue where Azure Communication Services (ACS) real-time transcription fails to start is usually related to configuration constraints rather than a code exception. Your code structure is generally correct, but there are a few important prerequisites and limitations to verify.

Key points to check:

Speech resource region must match ACS region. Real-time transcription requires the Speech service to be in the same Azure region as the ACS resource. If ACS is in eastus, Speech must also be in eastus. Cross-region Speech endpoints are not supported for call transcription. If the regions do not match, transcription will silently fail even though the call connects successfully.
- - Reference: Speech service supported regions
StartTranscription = true does not auto-start transcription In Call Automation, setting it only configures transcription capability at call connection time. It does not reliably start transcription for Room or connected calls. For real-time transcription, you must explicitly call StartTranscriptionAsync
- Reference: Start Transcription

Feel free to post back for further assistance.

Daniel Li 0 Reputation points

2025-12-16T23:25:34.76+00:00
Hi @Vimal Lalani

Thank you for your response and the detailed explanation. I have already verified the two points you mentioned:

Region alignment – My Speech resource is deployed in the same region as my ACS resource (no cross‑region setup).

StartTranscriptionAsync – I tried to invoke StartTranscriptionAsync after the call connection, but the transcription status still remains inactive.
// also tried with StartTranscriptionOptions try { CallAutomationClient callAutomationClient = new CallAutomationClient(this.connectionString); await callAutomationClient.GetCallConnection(callConnectionId).GetCallMedia().StartTranscriptionAsync(); }......

I also noticed something in the official documentation:

In the examples, the property previously referencing Cognitive Services now appears to be replaced by Speech service.

This makes me wonder if there might have been a change in the SDK or platform upgrade where ACS transcription support was not fully updated to align with the Speech service endpoint.

Could you please confirm whether ACS real-time transcription is currently working and requires a Speech resource instead of a Cognitive Services resource, and if there are any known gaps in support during this transition?
Vimal Lalani 2,560 Reputation points Microsoft External Staff Moderator

2025-12-17T20:07:19.0733333+00:00
Hi @Daniel Li

Thanks for the extra details. To answer your questions......

Does ACS real-time transcription require a Speech resource instead of Cognitive Services?

Yes. ACS real-time transcription uses an Azure AI Speech resource. Although Speech is part of the broader Azure AI (formerly Cognitive Services) family, the current guidance is to explicitly create and configure a Speech resource and use its endpoint.

The ACS documentation and samples now reference the Speech service because it provides the Automatic Speech Recognition (ASR) capability required for transcription. There is no separate or new integration path, this is primarily a terminology and configuration clarification.

Is ACS real-time transcription currently working?

Yes. ACS real-time transcription is supported and working when properly configured using the Call Automation SDKs (C#, JavaScript, Java, Python). Key requirements include:

An ACS resource linked to a valid Azure AI Speech resource

Using the Speech resource’s custom subdomain endpoint (for example, https://<your-subdomain>.cognitiveservices.azure.com/)

Correct CallIntelligenceOptions and TranscriptionOptions configuration (locale, transport, etc.)

Known confusion during this transition

Some older samples and references still use generic Cognitive Services terminology or legacy endpoints. This can be misleading. If an incorrect endpoint is used (such as a generic Cognitive Services endpoint instead of the Speech resource subdomain), transcription metadata may be received, but the transcription status can remain inactive.

This behavior typically indicates a configuration issue rather than a service outage or an unsupported scenario.

Hope this helps!
Daniel Li 0 Reputation points

2025-12-19T02:57:54.4733333+00:00
Hi @Vimal Lalani ,

Thanks for following up. This was very helpful.

Based on the error I was seeing:

"resultInformation": { "code": 400, "subCode": 8565, "message": "Action failed due to a bad request to Cognitive Services. Please check your input parameters." }

I now realize that the issue is likely related to the endpoint configuration. I was using a region-based Cognitive Services endpoint, whereas ACS real-time transcription may require the Speech resource custom subdomain endpoint instead of the regional endpoint.

After reviewing the documentation here: https://learn.microsoft.com/en-ca/azure/ai-services/speech-service/speech-services-private-link?tabs=portal#create-a-custom-domain-name

I plan to update my configuration to use the custom domain and retry the integration.

Thanks again for the clarification and your help.
Daniel Li 0 Reputation points

2025-12-19T04:23:14.4933333+00:00

I found the root cause: I was using a region-based Cognitive Services endpoint, but ACS real-time transcription requires the Speech resource custom subdomain endpoint instead.

Appreciate the guidance, and hope this helps anyone facing the same issue!

Share via

Azure Communication Services real time transcript failed to start

1 answer

Your answer