Speech-to-Text V2 API documentation

[mamyraimov]

In list of supported languages model "long" available for Kazakh language (kk-kz) but in reality it's not working? 
It's docs problem? Or there gonna be new updates in  supported languages?

[lsolatorio]

Hi @mamyraimov, 

Welcome and thank you for reaching out to the community.

I understand that the document can be confusing at times, please let me share some thoughts around it.

Filtering the language "kk-KZ" and looking at the "Long" models, you'll notice that the only supported feature available is "Profanity filter". For example, if your transcripted audio file didn't have any profanity then it is most likely that the model will return no model adaptation was used for the transcription.

Hope this helps.

[JonSU]

To my understanding, the "Profanity filter" is merely a feature indicating whether a model supports this functionality. Furthermore, according to the Google Cloud UI, if a model in a specific region does not support a certain language, we are prevented from submitting transcription tasks in that language. However, we were able to submit our task successfully; it was only after the task began execution that an error was reported.

This is not an error in the documentation or UI; it must be a bug in the system – the only logical explanation.

Please help us clarify this issue.

Thank you.

[mamyraimov]

 No, it's doesn't.
The problem in this message 
"The requested model is currently not supported for language : kk-KZ."
Why it's available to create task for transcription, but then it shows this error?
Something wrong, API or documentation.

[JonSU]

I've encountered the same issue: the "long" & "short" models fails to operate with "zh-tw." After extensive trial and error, only the "chirp" model proved compatible with zh-tw. However, the "chirp" model does not support StreamRecognizer, so I have no choice but to abandon v2 and revert to the less efficient v1.

This does not match what is written in the documentation!!