a little confusion

Hey!
I have some confusion about the computing process. The input audio whose size is 112501KB, gets an output of 112486KB. Could you tell me the reason and the operations about the audio throughout the prediction?

Thank you 