Troubleshoot common issues

Available in Classic and VPC

You might run into the following problems when using CLOVA Speech. Find out causes and possible solutions.

Security issues

Are there any security issues with the files uploaded to Object Storage and the data created in CLOVA Speech?

Causes and solutions

NAVER Cloud Platform does not access to your data. For more information, see CLOVA Speech terms of service.

Degraded voice recognition accuracy

The accuracy of voice recognition is degraded.

Cause

If the file format is not supported, the accuracy may be degraded.

Solution

Check if the format of the voice file meets the format that CLOVA Speech supports. For more information, see Scope of support.

Increase domain limit

I cannot create any more domains.
I would like to increase the limit for the number of domains.

Causes and solutions

• If you are a corporate customer, request the domain limit increase in the Service Quota service.
• If you are an individual customer, request to Customer Support.

Speaker separation during live streaming

I am using CLOVA Speech gRPC to call APIs in live streaming, and the "speaker" (speaker separation) result is not created.

Cause

Only long sentence recognition supports speaker separation. In live streaming, speaker separation is not available.

Solution

As explained in the cause section, it is a normal operation and it does not need any action. To use the speaker separation feature, use the long sentence recognition API.

English recognition errors

When I request a recognition task in the long sentence recognition service, it does not recognize it if there are English words in the sentence.

Cause

When the recognition language is set to Korean, English may not be recognized.

Solution

• When requesting a recognition task in the builder, change the Language setting to Korean/English simultaneous recognition.
• When calling APIs, change params.lang to enko, then call. For more information, see Long sentence recognition.