I'm making an audiodataset for voice commands

Should I add some examples with noise to make a versatile model?