TTS: Receiving google.api_core.exceptions.InternalServerError: 500 INTERNAL: ErrorSpacePayload error

[javatwo]

Hi there,

This was working before and suddenly stopped giving the following error : -

processing mlpy.txt --> mlpy.wav
gs://bucketname/mlpy.wav projects/myproject/locations/mylocation
Traceback (most recent call last):
File "txt_to_wav.py", line 73, in <module>
result = operation.result(timeout=900) ## wait upto 15 mins
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Python\Python312\Lib\site-packages\google\api_core\future\polling.py", line 261, in result
raise self._exception
google.api_core.exceptions.InternalServerError: 500 INTERNAL: [type.googleapis.com/util.ErrorSpacePayload='SpeechErrorSpace::TTS_BACKEND_REQUEST_RPC_ERROR'] 13: INTERNAL: [type.googleapis.com/util.ErrorSpacePayload='SpeechErrorSpace::TTS_BACKEND_REQUEST_RPC_ERROR']

Any one seen this before? Any solution? What do we do? Thanks for your help.

[Poala_Tenorio]

The 500 Internal Server Error often indicates an issue on the server-side, which might not be directly controllable from your end. Temporary issues can occur on the service provider's end. Trying the same operation after some time (an hour or more) might resolve the problem if it's due to a transient issue.

Remember to handle exceptions in your code to capture and log such errors gracefully, providing helpful messages for debugging and informing users when necessary.

[adamsiwiec]

Is this really the only solution for this? Waiting an hour? What if we have an application in production that utilizes this TTS, do we tell the customer to wait an hour? I've been increasingly getting this error over the past couple days, I have retry logic in my code but it will fail on 5 consecutive retries, then miraculously work later. 

Google shouldn't advertise TTS as a production service if this is expected behavior.

[adamsiwiec]

Are there any updates here? We are a few weeks from releasing a product that uses Google TTS. Our only blocker in moving to production is this 500 level TTS error we are receiving from Google as a result of the "Temporary issues can occur on the service provider's end" mentioned by Poala above. 

If this doesn't get resolved we're unfortunately going to have to move away from Google TTS last minute. I'm a huge Google Cloud fan. I'll be really disappointed if I have to rewrite the code for my client before a production release due to a bug within a managed API. I'll be hesitant to use GCP APIs in the future if this is the expected behavior. 

Other than this 500 error, the API is great for Long Text Speech Synthesis and more straight forward/faster than the alternatives. However, with this temporary server issue that persists for ~5-10 minute periods or so, it decreases the quality of our user experience drastically. Essentially, our application transcribes text generated by an LLM and reads them to the user. If the user has receives an error which causes them to wait 5-10 minutes for a transcription, this significantly impacts our user experience.

[gcp-bb]

Indeed, I too am seeing this error now. It's fairly random, but sadly also common. It's not dependent on the input.