How to set SpeechSynthesizer timeout?

[valerauko]

Occasionally I observe the following error: CancellationErrorCode "ServiceTimeout" with message "USP error: timeout waiting for the first audio chunk" after roughly 40 seconds.

While this is unfortunate, my question is really how do I set a shorter timeout on the client side? When it doesn't time out, even if it's slow it's a few seconds at worst (and obviously much less until the "first audio chunk").

So I'd like to set a timeout much shorter than the observed 40sec and fail quick (and potentially retry)

Looking at the docs for SpeechSynthesizer and SpeechConfig I couldn't find any way so far.

Env

Java 21 with eclipse-temurin:21-jre-noble
SDK version com.microsoft.cognitiveservices.speech client-sdk 1.41.1

[BrianMouncer]

I'm not sure any of the TTS timeout values will help in this case.
You might want to try using pre-connect, so you can connect before actually needing the connection, as well as possibly handling any retry attempts.

This will also speed up the first audio chunk latency, because the network initialization and web socket upgrade will have already been done.

cognitive-services-speech-sdk/samples/java/android/SpeechSynthesis/app/src/main/java/com/microsoft/cognitiveservices/speech/samples/speechsynthesis/MainActivity.java

Line 183 in e8684f6

public void onPreConnectButtonClicked(View v) {

[valerauko]

Is the ServiceTimeout a connection timeout? I had the impression that it's timing out waiting for the model to generate the sound, not the network (it tends to happen with some models way more often than others)

[jamesw966]

I also hit this issue today and I'd like the ability to fail much faster (a few seconds if possible as I have a conversational AI use case). Python 3.12.7, SDK 1.41.1.

[github-actions]

This item has been open without activity for 19 days. Provide a comment on status and remove "update needed" label.