How to force language of ASR?

[elch10]

I have dataset with Uzbek audios. But sometimes model predicts as [tr] or even [ru]. How can I force model to make prediction as [uz] language?

[merki86]

I had the same problem, all of my audios were recognized as TR, UZ but not KK (the original one). I think there's no possibility to force the program to recognize audio as N language without editing its sources. And I also think this repo is no longer maintained.