Google Speech to Text latency issues

[Harry Stuart]

I am developing a real time voice application and am streaming audio to the Google Speech API, however, I am getting response times of around 2 seconds, this is still too great a delay. Is it possible to get the delay to below 500ms? How big an impact do voice models and expected key word parameters actually have on the transcription? What is the average latency for streaming transcription?

Thank you, Harry 

[George (Cloud Platform Support)]

Hello Harry, 

The response times depend on more than one factor, most of them directly related to the quality of the initial recording. To improve response times in your case, down from 2 seconds, you are encouraged to follow more closely the related recommendations on the "Best Practices" documentation page. 

If the above information does not fully address your particular situation, you are most welcome coming back with more detail. How did you create your audio file initially? How did you record sound, which encoding? Did you accurately describe the audio data sent with your request to the Speech-to-Text API. Ensuring that the RecognitionConfig for your request describes the correct sampleRateHertz, encoding, and languageCode will result in the most accurate transcription and billing for your request. A sample file would help us in reproducing your issue on our side.