microsoft / Cognitive-SpeakerRecognition-Windows

Windows SDK for the Microsoft Speaker Recognition API, part of Cognitive Services
https://www.microsoft.com/cognitive-services/en-us/speaker-recognition-api
Other
60 stars 62 forks source link

The request amount for streaming is doubled, readme.md is incorrect #8

Open twilightgod opened 7 years ago

twilightgod commented 7 years ago

I'm trying Speaker Recognition API, in Azure document, the API call limit is 20/min. I make a streaming identification application which is modified from streaming demo, I set the step to 3s, which will send 20 API calls based on readme.md. But I get error after 30s, it shows { "error": { "code": "RateLimitExceeded", "message": "Rate limit is exceeded. Try again later." } }.

Then I setup Fiddler to monitor traffic, I found the identification process actually includes two API calls, one is uploading WAV file and get an URL to the result operation, another one is getting real result from this URL.

This behavior doubles request amount, which made the minimum step 6 seconds, it's too large for streaming.

I think Azure should raise the API call limit or provide a pricing option to get higher limit.