Getting Exception with an error code: 0x38 (SPXERR_AUDIO_SYS_LIBRARY_NOT_FOUND) when calling speechsdk.SpeechSynthesizer

Question

Getting Exception with an error code: 0x38 (SPXERR_AUDIO_SYS_LIBRARY_NOT_FOUND) when calling speechsdk.SpeechSynthesizer

Anonymous

Hi, I'm trying to make text to speech implementation. The application works locally fine with windows 10, Python 3.7.9 But when deployed to Azure Webapp I get Internal Server Error (500) And from logs I see trace:

2023-04-17T10:34:47.616406949Z RuntimeError: Exception with error code: 2023-04-17T10:34:47.616410949Z [CALL STACK BEGIN] 2023-04-17T10:34:47.616414949Z 2023-04-17T10:34:47.616418649Z /tmp/8db3f2c6c18ea77/antenv/lib/python3.7/site-packages/azure/cognitiveservices/speech/libMicrosoft.CognitiveServices.Speech.core.so(+0x1aed40) [0x7701c07fad40] 2023-04-17T10:34:47.616422949Z /tmp/8db3f2c6c18ea77/antenv/lib/python3.7/site-packages/azure/cognitiveservices/speech/libMicrosoft.CognitiveServices.Speech.core.so(+0x1f7d3c) [0x7701c0843d3c] 2023-04-17T10:34:47.616428049Z /tmp/8db3f2c6c18ea77/antenv/lib/python3.7/site-packages/azure/cognitiveservices/speech/libMicrosoft.CognitiveServices.Speech.core.so(synthesizer_create_speech_synthesizer_from_config+0xf3) [0x7701c071bcae] 2023-04-17T10:34:47.616432449Z /usr/lib/x86_64-linux-gnu/libffi.so.7(+0x6d1d) [0x7701c92ffd1d] 2023-04-17T10:34:47.616436349Z /usr/lib/x86_64-linux-gnu/libffi.so.7(+0x6289) [0x7701c92ff289] 2023-04-17T10:34:47.616440349Z /opt/python/3.7.16/lib/python3.7/lib-dynload/_ctypes.cpython-37m-x86_64-linux-gnu.so(_ctypes_callproc+0x2be) [0x7701c931409e] 2023-04-17T10:34:47.616444349Z /opt/python/3.7.16/lib/python3.7/lib-dynload/_ctypes.cpython-37m-x86_64-linux-gnu.so(+0xeaa8) [0x7701c9313aa8] 2023-04-17T10:34:47.616448449Z /opt/python/3.7.16/lib/libpython3.7m.so.1.0(PyObject_Call+0x79) [0x7701ca8d5999] 2023-04-17T10:34:47.616452349Z /opt/python/3.7.16/lib/libpython3.7m.so.1.0(_PyEval_EvalFrameDefault+0x1df9) [0x7701ca8bb9f9] 2023-04-17T10:34:47.616456449Z /opt/python/3.7.16/lib/python3.7/site-packages/debugpy/_vendored/pydevd/_pydevd_frame_eval/pydevd_frame_evaluator.cpython-37m-x86_64-linux-gnu.so(+0x18876) [0x7701c8dd0876] 2023-04-17T10:34:47.616460549Z /opt/python/3.7.16/lib/libpython3.7m.so.1.0(_PyEval_EvalCodeWithName+0x2ef) [0x7701ca8b8a8f] 2023-04-17T10:34:47.616464749Z /opt/python/3.7.16/lib/libpython3.7m.so.1.0(_PyFunction_FastCallDict+0x2ce) [0x7701ca8cf64e] 2023-04-17T10:34:47.616469049Z /opt/python/3.7.16/lib/libpython3.7m.so.1.0(_PyEval_EvalFrameDefault+0x1df9) [0x7701ca8bb9f9] 2023-04-17T10:34:47.616473249Z /opt/python/3.7.16/lib/python3.7/site-packages/debugpy/_vendored/pydevd/_pydevd_frame_eval/pydevd_frame_evaluator.cpython-37m-x86_64-linux-gnu.so(+0x18876) [0x7701c8dd0876] 2023-04-17T10:34:47.616477349Z /opt/python/3.7.16/lib/libpython3.7m.so.1.0(_PyEval_EvalCodeWithName+0x2ef) [0x7701ca8b8a8f] 2023-04-17T10:34:47.616481449Z /opt/python/3.7.16/lib/libpython3.7m.so.1.0(_PyFunction_FastCallDict+0x2ce) [0x7701ca8cf64e] 2023-04-17T10:34:47.616485349Z /opt/python/3.7.16/lib/libpython3.7m.so.1.0(_PyObject_Call_Prepend+0x5f) [0x7701ca8d519f] 2023-04-17T10:34:47.616489349Z [CALL STACK END] 2023-04-17T10:34:47.616493249Z 2023-04-17T10:34:47.616496949Z Exception with an error code: 0x38 (SPXERR_AUDIO_SYS_LIBRARY_NOT_FOUND) 2023-04-17T10:34:47.618473457Z 169.254.130.4 - - [17/Apr/2023 10:34:47] "POST /generate_audio HTTP/1.1" 500 - 2023-04-17T10:34:47.618495457Z INFO:werkzeug:169.254.130.4 - - [17/Apr/2023 10:34:47] "[35m[1mPOST /generate_audio HTTP/1.1[0m" 500 -

1 answer

Your answer

Answer 1

romungi-MSFT 48,916 Microsoft Employee Moderator

@Anonymous The most likely reason for this error is that on a webapp or app service you will have to set the audio config for TTS or STT scenarios since the remote app service does not have access to a microphone or speaker the audio config should be set to a audio file.

Please this [thread](https://learn.microsoft.com/en-us/answers/questions/1184428/azure-text-to-speech-error-code-0x38-(spxerr-audio) for details which fixed the issue for the user.

Try to add the following to your speechSyntheszier() config.

file_name = "outputaudio.wav"
    file_config = speechsdk.audio.AudioOutputConfig(filename=file_name)
    speech_synthesizer = speechsdk.SpeechSynthesizer(speech_config=speech_config, audio_config=file_config)

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.

Anonymous

2023-04-17T15:53:08.12+00:00
I do not understand this solution, there is a audio file created instead of this code:

audio_config = speechsdk.audio.AudioOutputConfig(use_default_speaker=True)

I did test this, after the change no error is present anymore, but neither is any sound played locally or in Azure Webapp. So unless you are suggesting I use playsound module or something afterwards this is not solution for my issue
romungi-MSFT 48,916 Reputation points Microsoft Employee Moderator

2023-04-17T16:44:15.0533333+00:00

With the error reported earlier, the default speaker cannot be enabled on a remote machine that is running the azure app service. Creating an audio file by setting the audio config to write to a file is required to download or play the file locally on your device or application. If the fie generated did not play locally, then the SDK logs should help understand if there is any issue while synthesizing.
Anonymous

2023-04-17T16:51:12.2766667+00:00

Yes a file is generated, but indeed the file does not play. What do you mean by SDK logs? No log is getting written now.
romungi-MSFT 48,916 Reputation points Microsoft Employee Moderator

2023-04-18T04:31:51.19+00:00
To enable logging with SDK, you can add the following to the speechConfig property.

speech_config.set_property(speechsdk.PropertyId.Speech_LogFilename, "LogfilePathAndName")

Since the file did not play, I think there could be an issue with some missing libraries. You might also want to check the similar setup with audio config on your local machine to check if the audio file generation works.
Deleted

This comment has been deleted due to a violation of our Code of Conduct. The comment was manually reported or identified through automated detection before action was taken. Please refer to our Code of Conduct for more information.
Anonymous

2023-04-18T09:29:20.63+00:00

Also I found this issue from github which is still open:

Internal work item ref. 3721132.

https://github.com/Azure-Samples/cognitive-services-speech-sdk/issues/1210
romungi-MSFT 48,916 Reputation points Microsoft Employee Moderator

2023-04-18T10:55:50.0233333+00:00

JS As per the log it seems the synthesis is complete but I am not sure why the audio does not contain the text you have provided in your SSML.

[88276]: 7673ms SPX_TRACE_VERBOSE: synthesizer.cpp:322 CSpxSynthesizer::ExecuteSynthesis: audio duration: 1387 ms.

I think the following text should be available as audio. testi

Did you try the similar scenario on your local machine with the same audio config? It might be helpful if you can also provide your code snippet to understand if any other config is missing.

The GH issue mentioned above seems to be similar but for a scenario where google collab was used as a platform to synthesize. You can go ahead and create an issue with SDK repo to work with SDK team but I think if the audio config of using a wav file works locally it should work on your app service.

Anonymous

Hi, Here is a simplified version what I'm working on: (sound is not played, but file is generated) and I can play the sound file with a media player and it is okay)

def generate_audio():

    translated_text = "testi"

    if not translated_text:
        return redirect(url_for('application'))

    target_lang = 'en'
    voice = 'en-US-JessaNeural' # Default to English (US)

    speech_config = speechsdk.SpeechConfig(subscription=os.environ.get('SPEECH_KEY'), region=region)
    speech_config.speech_synthesis_voice_name = voice
    speech_config.set_property(speechsdk.PropertyId.Speech_LogFilename, "LogFile")
  
    file_name = "audiofile.wav"
    file_config = speechsdk.audio.AudioOutputConfig(filename=file_name)
    
    synthesizer = speechsdk.SpeechSynthesizer(speech_config=speech_config, audio_config=file_config)
    result = synthesizer.speak_text_async(translated_text).get()

    if result.reason == speechsdk.ResultReason.SynthesizingAudioCompleted:
        audio_data = base64.b64encode(result.audio_data).decode()
        audio_url = "data:audio/mp3;base64," + audio_data
    else:
        logging.error("Error synthesizing audio: %s", result)
        print("Error synthesizing audio: {}".format(result))

    return redirect(url_for('application'))

And if change

file_name = "audiofile.wav"
file_config = speechsdk.audio.AudioOutputConfig(filename=file_name)
    
    synthesizer = speechsdk.SpeechSynthesizer(speech_config=speech_config, audio_config=file_config)

to


 audio_config = speechsdk.audio.AudioOutputConfig(use_default_speaker=True)
 synthesizer = speechsdk.SpeechSynthesizer(speech_config=speech_config, audio_config=audio_config)

It works locally, but I get the error Exception with an error code: 0x38 (SPXERR_AUDIO_SYS_LIBRARY_NOT_FOUND) in Azure Webapp

romungi-MSFT 48,916 Reputation points Microsoft Employee Moderator

2023-04-19T06:25:11.3133333+00:00

JS Thanks for sharing. With the audio file in the config, the file should play on any supported player for the for the file format. With respect to using a default speaker with App service, the app service does not have a default speaker, so the SDK throws an error with this config. It will work on your local machine or any device that supports a local speaker. I hope this helps!!
Anonymous

2023-04-19T07:00:02.47+00:00

I noticed that I might have left out a key information. So what I am doing is I have a webpage, where I have a button to generate the sound. And I want the audio to play in the browser. And locally it works as described. But not azure and I get the error. With the file I do not get sound at all still
romungi-MSFT 48,916 Reputation points Microsoft Employee Moderator

2023-04-19T15:44:45.53+00:00
I'll try to list the scenarios for clarity.

AudioConfig -> Local Speaker

Locally: Works

Azure App service: Errors Out

The above is an expected behavior since app service does not support a speaker.

AudioConfig -> Wav File

Locally: File gets created and voice is heard when played with media player

Azure App Service: File gets created but no voice is heard when file is downloaded and played locally.

With the second AudioConfig is this the behavior you are seeing with App service?

If yes, then what is the platform used for the app service? I think for linux some dependencies need to be added with certificates and libraries as mentioned here.

If the dependency is met (windows or linux) then file generated should contain audio if you download and play it.

If the voice is still not available in the audio then I would request to collect the SDK log and post an issue with SDK team on this repo to check for any discrepancies of this request.

Share via

Getting Exception with an error code: 0x38 (SPXERR_AUDIO_SYS_LIBRARY_NOT_FOUND) when calling speechsdk.SpeechSynthesizer

1 answer

I did test this, after the change no error is present anymore, but neither is any sound played locally or in Azure Webapp. So unless you are suggesting I use playsound module or something afterwards this is not solution for my issue

Your answer