how was the spectogram constructed in speech command tfjs model

Question

I am trying to perform offline recognition using the speech-command tfjs library. however, I have little information on the required params or how was the spectogram constr