Azure AI Speech
An Azure service that integrates speech processing into apps and services.
2,005 questions
Azure Speech To Text is returning empty text since last few days even when audio file has words.
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(137.6, 10%, 23%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EAP%3C/text%3E%3C/svg%3E)
arvind prajapati
5
Reputation points
I've integrated Azure speech-to-text Golang SDK into my project. I'm using go v1.16.6 buster in the Docker container. In last few days when I've added a new package which is kagome for processing Japanese text, now I'm facing an issue where Azure STT is returning empty text.
I don't know the reason for not processing the speech and returning empty text. please help ASAP because my application is live for the public and totally depends on Azure STT.
sharing screenshots of my go.mod file from git. Where you can see if any package affected it
Azure AI Speech
Azure AI services
Azure AI services
A group of Azure services, SDKs, and APIs designed to make apps more intelligent, engaging, and discoverable.
3,449 questions
{count} votes
-
arvind prajapati 5 Reputation points
2023-05-30T07:03:14.3766667+00:00 @romungi-MSFT can you help with it?
-
romungi-MSFT 48,836 Reputation points Microsoft Employee Moderator2023-05-30T16:49:59.0733333+00:00 @arvind prajapati I do not have much experience with golang but it seems like it could be a compatibility issue with the package you are using and the go version that Azure speech SDK supports.
go.mod of kagome project indicates go 1.19 is used but go speech sdk go.mod file uses 1.13 as per this file.
I think it would be easier to check with SDK team if the speech SDK is compatible with kagome project. Could you also try to raise an issue on this repo?
With respect to the service, do you also see no respone to the speech if you use the STT service through other means like the speech API, speech studio, etc?
-
arvind prajapati 5 Reputation points
2023-05-31T05:25:11.9166667+00:00 @romungi-MSFT I also thought Golang version would be the issue, but later I tested it on the Real-time Speech-To-Text tool using basic and custom models, then it give an error for audio files, which was quota exceeded.
I viewed the Matrics on speech service, which was showing that 16k seconds were used this month. But the monthly quota for F0(free) tier is 5h, right?
If the quota is exceeded, then why STT was transcribing some files? It should not process all files.
-
romungi-MSFT 48,836 Reputation points Microsoft Employee Moderator
2023-05-31T09:40:48.2233333+00:00 @arvind prajapati I think 16k seconds should roughly be around 4.5 hours and since this is from metrics there will be a delay to update the same. If files that are below the limit, then I think the service might process those and if it exceeds it might hit the quota error. You might want to switch to standard tier and check if your application works with the file. Thanks!!
-
arvind prajapati 5 Reputation points
2023-05-31T09:58:48.9133333+00:00 @romungi-MSFT The quota is still not exceeded, it's 16.9 k and still it's giving the quota exceeded error, as you can see in this screenshot of the real-time tool some records are processed, and some showing errors. This screenshot is recently taken.
-
romungi-MSFT 48,836 Reputation points Microsoft Employee Moderator
2023-05-31T10:14:20.68+00:00 @arvind prajapati Try switching to standard tier since free tier has limitations with TPS and other quotas. If you see persistent issues with standard tier you can always raise a support case to increase the limits further to ensure the endpoint can handle higher limits.
-
arvind prajapati 5 Reputation points
2023-05-31T10:44:19.84+00:00 @romungi-MSFT Okay, I'll tell the client about that. Thank you for the quick response.
-
arvind prajapati 5 Reputation points
2023-08-09T06:56:13.41+00:00 @romungi-MSFT @VasaviLankipalle-MSFT I switched to S0 standard tier yesterday at 05:30 PM IST. But sometimes the audio file gets transcribed and sometimes returns the Quota Exceed error. Why does the speech service return the error even after switching to the S0 Standard tier?
When I checked the Cost Analysis in my speech service, the cost for transcription was added to my account.
And if the cost is being added, then why does it return the Quota Exceeded error?- This image is of Realtime Speech To Text tool,
- Selected Price Tier
- Cost Analysis
- This image is of Realtime Speech To Text tool,
-
romungi-MSFT 48,836 Reputation points Microsoft Employee Moderator
2023-08-13T15:17:26.9833333+00:00 @arvind prajapati I think it is most likely that the service perceives sharp changes in the workload that requires the service to scale the capacity to accommodate the requests and it returns a quota exceeded error.
Even though you might be passing requests within the limits the backend service still needs to scale based on initial request rate as it evaluates the rate of incoming requests over a small period of time, typically 1 or 10 seconds and then determines if the rate limits are being exceeded. If it estimates that the rate could exceed error 429 is reported.
The general best practice in this case is to implement retry logic to resend the request to the API after few seconds.
Sign in to comment
Sign in to answer