Problem creating SpeechRecognizer with audio stream input using node.js Speech SDK

Question

Problem creating SpeechRecognizer with audio stream input using node.js Speech SDK

Michael Pickering 0

Using Speech SDK for JavaScript v1.44.0, and following the STT in-memory streaming example, but using the fromEndpoint API to create Recognizer, as recommended in the Release Notes for that SDK version. Node.js is v22 LTS, running in Azure Cloud as an Azure Web App. Azure Speech Service is in France Central.

Describe the bug

Call to create the SpeechRecognizer with the following error information:

**/node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/src/sdk/Audio/AudioConfig.js:184 return this.privSource.id(); ^

TypeError: this.privSource.id is not a function at AudioConfigImpl.id (/node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/src/sdk/Audio/AudioConfig.js:184:32) at new ServiceRecognizerBase (/node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/src/common.speech/ServiceRecognizerBase.js:63:88) at new SpeechServiceRecognizer (/node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/src/common.speech/SpeechServiceRecognizer.js:11:9) at SpeechRecognizer.createServiceRecognizer (/node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/src/sdk/SpeechRecognizer.js:216:16) at SpeechRecognizer.implCommonRecognizerSetup (/node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/src/sdk/Recognizer.js:102:30) at new Recognizer (/node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/src/sdk/Recognizer.js:26:14) at new SpeechRecognizer (/node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/src/sdk/SpeechRecognizer.js:26:9) at WebSocket. (/home/site/wwwroot/server.js:384:30) at WebSocket.emit (node:events:518:28) at Receiver.receiverOnMessage (/node_modules/ws/lib/websocket.js:1220:20)

Node.js v22.15.0**

To Reproduce

Create a node.js server in Azure Web App, create an Audio pushStream, then attempt to create a new SpeechRecognizer after a WebSocket client connects, like this:

  let pushStream = sdkSpeech.AudioInputStream.createPushStream(audioStreamFormat);

  // Get the audioConfig from STT Stream
  let audioConfig = sdkSpeech.AudioConfig.fromStreamInput(pushStream);

  // Prepare the recognizer
  let recognizeSTT = new sdkSpeech.SpeechRecognizer(speechConfig, audioConfig);

Expected behavior

Speech Recognizer is created and ready to recognize speech from the audio stream!

Version of the Cognitive Services Speech SDK

1.44.0

Platform, Operating System, and Programming Language

OS: Linux in Azure Web App with node.js 22 LTS Hardware - Whatever Azure is Using! Programming language: JavaScript (node.js) Browser - not applicable Additional context

The error crashes the node.js server, and the crash occurs AS SOON AS the call to create the SpeechRecognizer is sent and BEFORE sending any audio data to the STT service!

Version

1.36.0 (Latest)

What browser/platform are you seeing the problem on?

Node

Relevant log output

/node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/src/sdk/Audio/AudioConfig.js:184

1 answer

Your answer

Answer 1

Suwarna S Kale 3,951

Hello Michael Pickering,

Thank you for posting your question in the Microsoft Q&A forum.

The error you are experiencing occurs due to incompatibility between Speech SDK v1.44.0 and Node.js v22 in Azure Web Apps, specifically when initializing SpeechRecognizer with push streams. This appears to be a regression bug affecting Linux environments.

Recommended solutions:

Downgrade to SDK v1.36.0 for stability
Test with fromDefaultMicrophoneInput() as temporary workaround
Consider Windows-based App Service or container deployment

Next steps:

Report the bug via Microsoft's GitHub issues
Monitor for SDK patches in upcoming releases
REST API may serve as interim solution

The Cognitive Services team is aware of similar streaming issues and working on a resolution. For production systems, SDK v1.36.0 remains the most reliable option currently.

If the above answer helped, please do not forget to "Accept Answer" as this may help other community members to refer the info if facing a similar issue. Your contribution to the Microsoft Q&A community is highly appreciated.

Michael Pickering 0 Reputation points

2025-06-05T18:44:24.3266667+00:00

Thanks, I'll try the suggested SDK version downgrade soon! Meanwhile, could you please expand on the suggested workaround using fromDefaultMicrophoneInput() ? I didn't think that would be an option in node.js. Do you mean just to use it to create the Recognizer, but then push audio like I would with pushStream? Can you show some example code please?
Ravada Shivaprasad 920 Reputation points Microsoft External Staff Moderator

2025-06-09T21:25:07.8533333+00:00

Hi Michael Pickering

You're correct that fromDefaultMicrophoneInput() is not supported in Node.js environments—it is intended for use in browser-based applications where microphone access is available. The suggestion to use it was likely meant for local testing purposes only, to help isolate whether the issue lies specifically with PushAudioInputStream in SDK v1.44.0. In Node.js, the standard approach remains using PushAudioInputStream with audio data from a file or other stream source. For example, you can create a recognizer using a push stream like this:

For now, downgrading to SDK v1.36.0 is the most stable solution for production use in Node.js on Azure Web Apps. Let me know if you'd like help drafting a GitHub issue to report this regression.

Thank you.
Ravada Shivaprasad 920 Reputation points Microsoft External Staff Moderator

2025-06-10T20:47:12.0333333+00:00

Hi Michael Pickering

Following up to see if the above answer was helpful. If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know

Thanks
Michael Pickering 0 Reputation points

2025-06-29T13:33:55.1366667+00:00

This is helpful, but for production a real fix to the SDK will be needed!
Michael Pickering 0 Reputation points

2025-06-29T13:44:03.5866667+00:00

Reverted to the suggested version, v1.36.0, of the MS Speech SDK, but got the same error message!
Pavankumar Purilla 10,350 Reputation points Microsoft External Staff Moderator

2025-07-02T05:13:25.8566667+00:00

Hi Michael Pickering,
Thank you for your update. Based on your observation that reverting to version 1.36.0 of the Microsoft Speech SDK still results in the same error, it appears that the issue is not limited to a specific SDK version. The root cause likely lies in how the PushAudioInputStream is being used together with AudioConfig.fromStreamInput() and SpeechRecognizer. The SDK internally expects the audio stream object to have a specific structure, including an id() method, which the PushAudioInputStream does not expose in the expected context.

This issue can be further complicated when using SpeechConfig.fromEndpoint(), as this method is not fully compatible with streaming audio inputs like push streams. To move forward, we recommend using SpeechConfig.fromSubscription() or SpeechConfig.fromAuthorizationToken() instead of fromEndpoint() when working with streaming audio input.

Please also verify that the push stream is created using sdk.AudioInputStream.createPushStream() and passed directly to AudioConfig.fromStreamInput() without any additional modifications.

If the issue continues, it might be related to the Node.js v22 runtime environment in your Azure Web App. In this case, we suggest testing the same code in a local environment or using a stable LTS version such as Node.js v18 to help isolate whether the runtime environment is contributing to the problem.
Michael Pickering 0 Reputation points

2025-07-02T08:37:10.06+00:00
Thanks for this information!

Regarding the node.js runtime, in fact node.js v22 IS the current, stable LTS version, and has been for some time. Furthermore, node.js v18 will be officially out of support on Azure as of July 10, 2025, which is in about one week!

Nevertheless, I already went ahead and tested my code with both the current Speech SDK version (1.44.1) and the recommended older, stable version (1.36.0) on all the currently deployable to Azure Web App node.js versions, namely, 18, 20, and 22. The same error and message was the result in each case!

Also, according to the release notes for Speech SDK for JS v 1.44.0, the "FromEndpoint" API is the recommended method for constructing a SpeechRecognizer. Please see here:
|https://github.com/Microsoft/cognitive-services-speech-sdk-js/releases

In particular, I call your attention to this section:

Updated FromEndpoint API to be the recommended method for constructing a SpeechConfig for most scenarios.

Applies to using:

SpeechRecognizer

TranslationRecognizer (via SpeechTranslationConfig)

ConversationTranscriber

SpeechSynthesizer

You can now use the Endpoint from the Azure Portal for Speech and Cognitive Services resources to construct a SpeechConfig object.
Michael Pickering 0 Reputation points

2025-07-02T08:42:00.6533333+00:00

I will go ahead and work on coding and testing another workaround using the SpeechConfig.fromSubscription method. Meanwhile, it would be very helpful to update the SDK release notes to reflect that FromEndpoint does not yet work in all scenarios!
Michael Pickering 0 Reputation points

2025-07-02T10:43:23.4933333+00:00

I just modified my code to use fromSubscription to create the Speech configuration. Unfortunately, I got the same error & message again! In this case I was using the latest SDK release again, should I try also reverting to the older SDK release. namely, 1.36.0?
Ravada Shivaprasad 920 Reputation points Microsoft External Staff Moderator

2025-07-21T23:43:35.9633333+00:00

Hi Michael Pickering

You’re on the right track by switching to SpeechConfig.fromSubscription, as fromEndpoint is known to have limitations with streaming audio inputs like PushAudioInputStream. However, the error you're encountering is likely due to how the stream is being handled rather than the SDK version itself. Downgrading to version 1.36.0 may not resolve the issue. Instead, ensure you're using AudioInputStream.createPushStream() correctly and passing it directly to AudioConfig.fromStreamInput(). Also, it would be beneficial for the SDK release notes to clarify that fromEndpoint doesn’t support all scenarios, especially with push streams.

Thank you
Ravada Shivaprasad 920 Reputation points Microsoft External Staff Moderator

2025-07-27T08:12:24.0866667+00:00

Hi Michael Pickering

Did you get any chance to check the response.

Thank you!

Share via

Problem creating SpeechRecognizer with audio stream input using node.js Speech SDK

1 answer

Your answer