I remember seeing some video where they did a FT of an audio sample and then just used mspaint to remove some frequency component and transformed back to the audio / time domain.

Something along those lines anyway.