In speech recognition when feeding mfcc coefficients(13) to bpnn(ANN) should all the input .wav or audios should have same number of feature vectors i.e each wav file has many frames and get many of 13Mfcc's

In speech recognition when feeding mfcc coefficients(13) to bpnn(ANN) should all the input .wav or audios should have same number of feature vectors i.e each wav file has many frames and get many of 13Mfcc's i.e all input .wav after collecting the 13-mfcc when fed for each .wav many of frames i.e many of 13-mfccs then question is should all .wav's should have same number of 13-mfccs (25ms)