C# Azure SpeechSynthesizer Timeout

Burning Aurora 0 Reputation points
2024-06-23T15:20:19.51+00:00

I recently started running into a "Timeout while synthesizing" error when trying to use the OpenAI TTS Voices. Coming from the northcentralus region on a S0 plan. It will sporadically work, but am not seeing anything in my logs about rate limits. Even the simplest 2 word sentence will fail. It has been working for weeks up until a few days ago. I'm really scratching my head here so any assistance would be great. A "dumbed" down version of the code I'm running is below.

Using the sample code from SpeechStudio without making any changes to what is provided I get "Connection was closed by the remote host. Error code: 1013. Error details: Status(StatusCode="ResourceExhausted", Detail="AOAI Resource Exhausted") USP state: Sending. Received audio size: 0 bytes."

 			// To support Chinese Characters on Windows platform
           if (Environment.OSVersion.Platform == PlatformID.Win32NT)
           {
               Console.InputEncoding = System.Text.Encoding.Unicode;
               Console.OutputEncoding = System.Text.Encoding.Unicode;
           }
 
           // Creates an instance of a speech config with specified subscription key and service region.
           var config = SpeechConfig.FromSubscription("subscriptionKey", "northcentralus");
 
           // Creates a speech synthesizer using the default speaker as audio output.
           using var fileOutput = Microsoft.CognitiveServices.Speech.Audio.AudioConfig.FromWavFileOutput(@"C:\temp\TestFile.wav");
           using var synthesizer = new SpeechSynthesizer(config, fileOutput);
           // Sets the synthesis output format.
           // The full list of supported format can be found here:
           // https://docs.microsoft.com/azure/cognitive-services/speech-service/rest-text-to-speech#audio-outputs
           //config.SetSpeechSynthesisOutputFormat(SpeechSynthesisOutputFormat.Audio16Khz128KBitRateMonoMp3);
 
           // Receive a text from console input and synthesize it to speaker.
           using var result = await synthesizer.SpeakSsmlAsync("[pause] Test Output");
           if (result.Reason == ResultReason.SynthesizingAudioCompleted)
               Console.WriteLine($"Successfully created {fileName}");
           else if (result.Reason == ResultReason.Canceled)
           {
               var cancellation = SpeechSynthesisCancellationDetails.FromResult(result);
               var error = new StringBuilder();
               error.AppendLine($"Error creating {fileName}{Environment.NewLine}");
               error.AppendLine($"CANCELED: Reason={cancellation.Reason}");
 
               if (cancellation.Reason == CancellationReason.Error && cancellation.ErrorCode != CancellationErrorCode.TooManyRequests)
               {
                   error.AppendLine($"CANCELED: ErrorCode={cancellation.ErrorCode}");
                   error.AppendLine($"CANCELED: ErrorDetails=[{cancellation.ErrorDetails}]");
               }
               else if (cancellation.Reason == CancellationReason.Error && cancellation.ErrorCode == CancellationErrorCode.TooManyRequests)
               {
                   Console.WriteLine($"Sleeping for 1 minute");
                   Thread.Sleep(60000);
               }
               throw new Exception(error.ToString());
           }
Azure AI Speech
Azure AI Speech
An Azure service that integrates speech processing into apps and services.
1,778 questions
0 comments No comments
{count} votes

1 answer

Sort by: Most helpful
  1. Azar 22,950 Reputation points MVP
    2024-06-23T20:24:54.16+00:00

    Hi there Burning Aurora

    Thanks for using QandA platform

    I guess this issue is due to the duration limit 10 minutes , guess thats the reason why you are hitting the limits,

    You will have to use the long audio synthesize larger content. Please see the referenced page.

    https://learn.microsoft.com/en-in/azure/ai-services/speech-service/batch-synthesis

    If this helps kindly accept the answer thanks ,much,


Your answer

Answers can be marked as Accepted Answers by the question author, which helps users to know the answer solved the author's problem.